What problem does it solve?
Target prioritization and marker-gene discovery require knowing whether a gene's expression is restricted to specific cell types, but single-cell pipelines typically stop at embedding and clustering without returning a per-gene specificity metric. This Skill quantifies cell-type specificity from an annotated single-cell atlas using the tau specificity index and Sarle's bimodality coefficient.
Core Features & Use Cases
- Tau Specificity Index: Computes the Yanai et al. 2005 tau index over pseudobulk per-cell-type means, excluding cell types with fewer than 20 cells.
- Bimodality Coefficient: Calculates Sarle's bias-corrected BC over expressing cells to detect on/off expression patterns, with a low-expression flag for sparse genes.
- Ranked Cell Types & Trial Prior: Outputs ranked expressing cell types, a tidy per-cell-type CSV, and optionally attaches Zhang et al. 2026 trial-success odds ratios.
- Use Case: Given a lung atlas in h5ad format, profile CD276 to learn it is highly cell-type-specific (tau > 0.69), then feed the JSON profile into downstream target-validation scoring.
Quick Start
Ask the agent to run the celltype-specificity-profiler demo to profile the MS4A1 gene on the bundled real pbmc3k dataset and return its tau and bimodality scores.
